Authors :
Presenting Author: Zhengxiang Cai, PhD – Carnegie Mellon University
Vasileios Kokkinos, PhD – Epilepsy Surgery Neurophysiologist, Department of Neurology, Northwestern University; Boney Joseph, PhD – Post-Doctoral Fellow, Mayo Clinic; Shuai Ye, PhD – Department of Biomedical Engineering – Carnegie Mellon University; Alexandra Urban, MD – Neurologist, University of Pittsburgh Comprehensive Epilepsy Center, University of Pittsburgh School of Medicine; Anto Bagić, MD, PhD – Professor, Chief, And Director, University of Pittsburgh Comprehensive Epilepsy Center, University of Pittsburgh School of Medicine; Gregory Worrell, M.D., Ph.D – Neurologist, Department of Neurology, Mayo Clinic; Mark Richardson, MD, PhD – Neurosurgeon, Department of Neurology, Massachusetts General Hospital; Bin He, PhD – Trustee Professor of Biomedical Engineering, Biomedical Engineering, Carnegie Mellon University
Rationale:
Epilepsy, a prevalent neurological disorder, is challenging to treat. About one third of patients are resistant to medications. Surgical removal of pathological tissues can alleviate seizures if the epileptogenic zone (EZ) is accurately identified. High-frequency oscillations (HFOs) in intracranial EEG (iEEG) recordings hold promise for locating the EZ and guiding surgery. However, identifying and distinguishing HFOs is difficult, and established HFO measures show variable results. This study aims to develop an approach focusing on spatiotemporal sequencing of HFOs to identify pathological activities and correlate them with the EZ in focal epilepsy patients.
Methods:
We collected and analyzed pre-surgical magnetic resonance imaging (MRI) data and iEEG recordings from 11 patients with medically intractable epilepsy from two clinical centers. Individual head models and iEEG electrode positions were co-registered, and at least twelve hours of iEEG recordings were examined in each patient. The raw iEEG data were preprocessed to eliminate noisy channels with poor quality. Subsequently, our in-house algorithm was applied to identify all potential HFOs (including pathological and physiological, aHFOs) present in the iEEG. The detected events were extracted as multichannel epochs for sequencing analysis using our proposed approach. We estimated the underlying cortical distribution of HFO-sequences (sHFOs) for each patient and validated the results by comparing them to the seizure-onset zone (SOZ) determined by clinical experts. Additionally, we investigated the spatial distribution of all general HFOs detected for each patient to provide a comparative analysis.
Results:
In this cohort of 11 patients, robust detection of HFOs and identification of sequenced activities were achieved for all patients. Over 1 million putative HFOs were detected in total (50.8%±15.0% during the day, mean±SD, n=11), and about 72 thousand HFO-sequences were identified (42.0%±15.0% during the day, n=11). The number of sHFOs identified during the night was significantly higher than during the day (n=11; p=0.0186). Across all patients, the asymmetry rate of sHFOs was significantly higher compared to all detected HFOs (aHFOs: 0.66±0.61, sHFOs: 0.39±0.42; n=11; p=0.032). Moreover, the cortical distribution of sHFOs localized to the SOZ with an averaged localization error of 4.42±8.75 mm, demonstrating superior performance when compared to two other established measures: the overall HFO-rate and the leading HFO channels (n=11; HFO-lead: 14.93±11.49 mm, p=0.0039; aHFOs: 25.17±9.38 mm, p=0.0029). These findings suggest that HFO-sequences possess a stronger pathological association and serve as a refined biomarker in our tested cohort.
Conclusions:
In this study, we proposed and identified a refined biomarker for localization of epileptic brain through spatiotemporal sequencing of HFOs. The HFO-sequences showed a high association with clinical evidence and better pathological relevance than conventional HFO measures. This approach holds promise for studying epileptic brain activities and guiding surgical interventions during pre-surgical diagnosis.
Funding:
This work was supported in part by NIH R01 NS096761 and EB021027.